From: Ralph Campbell <rcampbell@nvidia.com> To: <nouveau@lists.freedesktop.org>, <linux-rdma@vger.kernel.org>, <linux-mm@kvack.org>, <linux-kselftest@vger.kernel.org>, <linux-kernel@vger.kernel.org> Cc: Jerome Glisse <jglisse@redhat.com>, John Hubbard <jhubbard@nvidia.com>, Christoph Hellwig <hch@lst.de>, Jason Gunthorpe <jgg@mellanox.com>, "Ben Skeggs" <bskeggs@redhat.com>, Andrew Morton <akpm@linux-foundation.org>, Shuah Khan <shuah@kernel.org>, Ralph Campbell <rcampbell@nvidia.com> Subject: [PATCH 03/16] nouveau: fix mixed normal and device private page migration Date: Fri, 19 Jun 2020 14:56:36 -0700 [thread overview] Message-ID: <20200619215649.32297-4-rcampbell@nvidia.com> (raw) In-Reply-To: <20200619215649.32297-1-rcampbell@nvidia.com> The OpenCL function clEnqueueSVMMigrateMem(), without any flags, will migrate memory in the given address range to device private memory. The source pages might already have been migrated to device private memory. In that case, the source struct page is not checked to see if it is a device private page and incorrectly computes the GPU's physical address of local memory leading to data corruption. Fix this by checking the source struct page and computing the correct physical address. Signed-off-by: Ralph Campbell <rcampbell@nvidia.com> --- drivers/gpu/drm/nouveau/nouveau_dmem.c | 8 ++++++++ 1 file changed, 8 insertions(+) diff --git a/drivers/gpu/drm/nouveau/nouveau_dmem.c b/drivers/gpu/drm/nouveau/nouveau_dmem.c index cc9993837508..f6a806ba3caa 100644 --- a/drivers/gpu/drm/nouveau/nouveau_dmem.c +++ b/drivers/gpu/drm/nouveau/nouveau_dmem.c @@ -540,6 +540,12 @@ static unsigned long nouveau_dmem_migrate_copy_one(struct nouveau_drm *drm, if (!(src & MIGRATE_PFN_MIGRATE)) goto out; + if (spage && is_device_private_page(spage)) { + paddr = nouveau_dmem_page_addr(spage); + *dma_addr = DMA_MAPPING_ERROR; + goto done; + } + dpage = nouveau_dmem_page_alloc_locked(drm); if (!dpage) goto out; @@ -560,6 +566,7 @@ static unsigned long nouveau_dmem_migrate_copy_one(struct nouveau_drm *drm, goto out_free_page; } +done: *pfn = NVIF_VMM_PFNMAP_V0_V | NVIF_VMM_PFNMAP_V0_VRAM | ((paddr >> PAGE_SHIFT) << NVIF_VMM_PFNMAP_V0_ADDR_SHIFT); if (src & MIGRATE_PFN_WRITE) @@ -615,6 +622,7 @@ nouveau_dmem_migrate_vma(struct nouveau_drm *drm, struct migrate_vma args = { .vma = vma, .start = start, + .src_owner = drm->dev, }; unsigned long i; u64 *pfns; -- 2.20.1
WARNING: multiple messages have this Message-ID (diff)
From: Ralph Campbell <rcampbell-DDmLM1+adcrQT0dZR+AlfA@public.gmane.org> To: nouveau-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org, linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, linux-mm-Bw31MaZKKs3YtjvyW6yDsg@public.gmane.org, linux-kselftest-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org Cc: Ralph Campbell <rcampbell-DDmLM1+adcrQT0dZR+AlfA@public.gmane.org>, Jason Gunthorpe <jgg-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>, Ben Skeggs <bskeggs-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>, Andrew Morton <akpm-de/tnXTf+JLsfHDXvbKv3WD2FQJk+8+b@public.gmane.org>, Shuah Khan <shuah-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org>, Christoph Hellwig <hch-jcswGhMUV9g@public.gmane.org> Subject: [PATCH 03/16] nouveau: fix mixed normal and device private page migration Date: Fri, 19 Jun 2020 14:56:36 -0700 [thread overview] Message-ID: <20200619215649.32297-4-rcampbell@nvidia.com> (raw) In-Reply-To: <20200619215649.32297-1-rcampbell-DDmLM1+adcrQT0dZR+AlfA@public.gmane.org> The OpenCL function clEnqueueSVMMigrateMem(), without any flags, will migrate memory in the given address range to device private memory. The source pages might already have been migrated to device private memory. In that case, the source struct page is not checked to see if it is a device private page and incorrectly computes the GPU's physical address of local memory leading to data corruption. Fix this by checking the source struct page and computing the correct physical address. Signed-off-by: Ralph Campbell <rcampbell-DDmLM1+adcrQT0dZR+AlfA@public.gmane.org> --- drivers/gpu/drm/nouveau/nouveau_dmem.c | 8 ++++++++ 1 file changed, 8 insertions(+) diff --git a/drivers/gpu/drm/nouveau/nouveau_dmem.c b/drivers/gpu/drm/nouveau/nouveau_dmem.c index cc9993837508..f6a806ba3caa 100644 --- a/drivers/gpu/drm/nouveau/nouveau_dmem.c +++ b/drivers/gpu/drm/nouveau/nouveau_dmem.c @@ -540,6 +540,12 @@ static unsigned long nouveau_dmem_migrate_copy_one(struct nouveau_drm *drm, if (!(src & MIGRATE_PFN_MIGRATE)) goto out; + if (spage && is_device_private_page(spage)) { + paddr = nouveau_dmem_page_addr(spage); + *dma_addr = DMA_MAPPING_ERROR; + goto done; + } + dpage = nouveau_dmem_page_alloc_locked(drm); if (!dpage) goto out; @@ -560,6 +566,7 @@ static unsigned long nouveau_dmem_migrate_copy_one(struct nouveau_drm *drm, goto out_free_page; } +done: *pfn = NVIF_VMM_PFNMAP_V0_V | NVIF_VMM_PFNMAP_V0_VRAM | ((paddr >> PAGE_SHIFT) << NVIF_VMM_PFNMAP_V0_ADDR_SHIFT); if (src & MIGRATE_PFN_WRITE) @@ -615,6 +622,7 @@ nouveau_dmem_migrate_vma(struct nouveau_drm *drm, struct migrate_vma args = { .vma = vma, .start = start, + .src_owner = drm->dev, }; unsigned long i; u64 *pfns; -- 2.20.1
next prev parent reply other threads:[~2020-06-19 21:58 UTC|newest] Thread overview: 79+ messages / expand[flat|nested] mbox.gz Atom feed top 2020-06-19 21:56 [PATCH 00/16] mm/hmm/nouveau: THP mapping and migration Ralph Campbell 2020-06-19 21:56 ` Ralph Campbell 2020-06-19 21:56 ` [PATCH 01/16] mm: fix migrate_vma_setup() src_owner and normal pages Ralph Campbell 2020-06-19 21:56 ` Ralph Campbell 2020-06-19 21:56 ` [PATCH 02/16] nouveau: fix migrate page regression Ralph Campbell 2020-06-19 21:56 ` Ralph Campbell 2020-06-19 21:56 ` Ralph Campbell [this message] 2020-06-19 21:56 ` [PATCH 03/16] nouveau: fix mixed normal and device private page migration Ralph Campbell 2020-06-19 21:56 ` [PATCH 04/16] mm/hmm: fix test timeout on slower machines Ralph Campbell 2020-06-19 21:56 ` Ralph Campbell 2020-06-19 21:56 ` [PATCH 05/16] mm/hmm/test: remove redundant page table invalidate Ralph Campbell 2020-06-19 21:56 ` Ralph Campbell 2020-06-19 21:56 ` [PATCH 06/16] mm/hmm: test mixed normal and device private migrations Ralph Campbell 2020-06-19 21:56 ` Ralph Campbell 2020-06-19 21:56 ` [PATCH 07/16] nouveau: make nvkm_vmm_ctor() and nvkm_mmu_ptp_get() static Ralph Campbell 2020-06-19 21:56 ` Ralph Campbell 2020-06-19 21:56 ` [PATCH 08/16] nouveau/hmm: fault one page at a time Ralph Campbell 2020-06-19 21:56 ` Ralph Campbell 2020-06-22 17:22 ` Jason Gunthorpe 2020-06-22 17:22 ` Jason Gunthorpe 2020-06-22 18:44 ` Ralph Campbell 2020-06-22 18:44 ` Ralph Campbell 2020-06-19 21:56 ` [PATCH 09/16] mm/hmm: add output flag for compound page mapping Ralph Campbell 2020-06-19 21:56 ` Ralph Campbell 2020-06-22 17:25 ` Jason Gunthorpe 2020-06-22 17:25 ` Jason Gunthorpe 2020-06-22 18:10 ` Ralph Campbell 2020-06-22 18:10 ` Ralph Campbell 2020-06-22 23:18 ` Jason Gunthorpe 2020-06-22 23:18 ` Jason Gunthorpe 2020-06-22 23:26 ` Ralph Campbell 2020-06-22 23:26 ` Ralph Campbell 2020-06-19 21:56 ` [PATCH 10/16] nouveau/hmm: support mapping large sysmem pages Ralph Campbell 2020-06-19 21:56 ` Ralph Campbell 2020-06-19 21:56 ` [PATCH 11/16] hmm: add tests for HMM_PFN_COMPOUND flag Ralph Campbell 2020-06-19 21:56 ` Ralph Campbell 2020-06-19 21:56 ` [PATCH 12/16] mm/hmm: optimize migrate_vma_setup() for holes Ralph Campbell 2020-06-19 21:56 ` Ralph Campbell 2020-06-19 21:56 ` [PATCH 13/16] mm: support THP migration to device private memory Ralph Campbell 2020-06-19 21:56 ` Ralph Campbell 2020-06-21 23:20 ` Zi Yan 2020-06-21 23:20 ` Zi Yan 2020-06-22 19:36 ` Ralph Campbell 2020-06-22 19:36 ` Ralph Campbell 2020-06-22 20:10 ` Zi Yan 2020-06-22 20:10 ` Zi Yan 2020-06-22 21:31 ` Ralph Campbell 2020-06-22 21:31 ` Ralph Campbell 2020-06-22 21:53 ` Zi Yan 2020-06-22 21:53 ` Zi Yan 2020-06-22 22:30 ` Yang Shi 2020-06-22 22:30 ` Yang Shi 2020-06-22 22:33 ` Yang Shi 2020-06-22 22:33 ` Yang Shi 2020-06-22 22:33 ` Yang Shi 2020-06-22 23:01 ` John Hubbard 2020-06-22 23:01 ` John Hubbard 2020-06-22 23:54 ` Yang Shi 2020-06-22 23:54 ` Yang Shi 2020-06-22 23:54 ` Yang Shi 2020-06-23 0:05 ` Ralph Campbell 2020-06-23 0:05 ` Ralph Campbell 2020-06-23 2:51 ` Huang, Ying 2020-06-23 2:51 ` Huang, Ying 2020-06-23 2:51 ` Huang, Ying 2020-06-19 21:56 ` [PATCH 14/16] mm/thp: add THP allocation helper Ralph Campbell 2020-06-19 21:56 ` Ralph Campbell 2020-06-22 0:15 ` Zi Yan 2020-06-22 0:15 ` Zi Yan 2020-06-22 21:33 ` Ralph Campbell 2020-06-22 21:33 ` Ralph Campbell 2020-06-19 21:56 ` [PATCH 15/16] mm/hmm/test: add self tests for THP migration Ralph Campbell 2020-06-19 21:56 ` Ralph Campbell 2020-06-19 21:56 ` [PATCH 16/16] nouveau: support THP migration to private memory Ralph Campbell 2020-06-19 21:56 ` Ralph Campbell 2020-06-22 12:39 ` [PATCH 00/16] mm/hmm/nouveau: THP mapping and migration Jason Gunthorpe 2020-06-22 12:39 ` Jason Gunthorpe 2020-06-22 16:58 ` Ralph Campbell 2020-06-22 16:58 ` Ralph Campbell
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=20200619215649.32297-4-rcampbell@nvidia.com \ --to=rcampbell@nvidia.com \ --cc=akpm@linux-foundation.org \ --cc=bskeggs@redhat.com \ --cc=hch@lst.de \ --cc=jgg@mellanox.com \ --cc=jglisse@redhat.com \ --cc=jhubbard@nvidia.com \ --cc=linux-kernel@vger.kernel.org \ --cc=linux-kselftest@vger.kernel.org \ --cc=linux-mm@kvack.org \ --cc=linux-rdma@vger.kernel.org \ --cc=nouveau@lists.freedesktop.org \ --cc=shuah@kernel.org \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: linkBe sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.